Application of a Modified Healthy Eating Index (HEI-Flex) to Compare the Diet Quality of Flexitarians, Vegans and Omnivores in Germany

Interest in plant-based nutrition has steadily increased in the western world in the recent years. The number of people following a meat-reduced, flexitarian diet is growing continuously. However, little is known about the diet quality of flexitarians compared to vegans or omnivores. Therefore, in this cross-sectional study, the food intake of 94 participants aged between 25–45 years was recorded via a validated food frequency questionnaire and 28 self-designed questions about the consumption of plant-based alternatives. An adapted Healthy Eating Index, HEI-flex, was developed to evaluate the diet quality of flexitarians, vegans and omnivores. Higher score points (SP) of the HEI-flex are associated with higher compliance with the official diet recommendations (Vmax = 100 SP). Finally, flexitarians scored significantly more highly when compared to omnivores (54 ± 8 vs. 47 ± 9 SP; p = 0.008) but lower than vegans (54 ± 8 vs. 61 ± 10 SP; p = 0.010). The results showed that the HEI-flex is a useful tool for assessing and comparing the diet quality of flexitarians, vegans and omnivores. Despite the consumption of highly processed plant-based alternatives, reduction in meat and meat products seems to be accompanied by increased overall diet quality.


Introduction
Plant-based diets are becoming increasingly popular in the Western world [1][2][3]. Although vegans (V) do not eat any food of animal origin, the food choices of vegetarians are very heterogeneous. Certain animal products are included or excluded depending on the form of vegetarianism. Additionally, there is a rising prevalence of the flexitarian (FX) diet in Western countries, which is broadly characterized by a primarily vegetarian diet pattern with occasional meat or fish consumption [4][5][6][7]. However, a generally accepted definition of flexitarianism does not currently exist. Instead, different definitions and consumption quantities of meat and meat products are discussed in various studies [8][9][10]. In Germany, the recommendations of the German Nutrition Society (DGE) define an FX diet as consuming meat or fish fewer than 4 days per week [11,12].
Interestingly, despite the rising popularity of a meat-reduced FX diet, annual meat consumption in Germany is currently only slightly declining, while the consumption of plant-based meat alternatives is growing constantly. The sales of plant-based meat alternatives overall increased by over 30% in 2021 compared to the previous year [13][14][15][16]. A consumer survey (n = 1000) in 2020 reported that half of the participants had purchased vegetarian or V meat alternatives in the past. Furthermore, FX households were four times more likely to consume vegetarian or V meat alternatives than omnivorous households. Plant-based alternatives, such as vegetarian/V burger patties, are suggested to be an equivalent (or better) nutritional substitute to the animal-based, conventional products in advertising [5,13,[17][18][19][20][21][22][23][24].
A cross-sectional pilot study including 94 participants was conducted from January to June 2020 using recruitment by flyers, social media and online communities.
The eligibility of participants was assessed using an online questionnaire to define their dietary patterns. This self-defined diet was confirmed during a subsequent telephone interview specifically focused on the daily intake of meat and meat-derivates to accurately differentiate between FX and OMN.
The present study included either participants who (a) maintained a flexitarian (FX) diet (i.e., plant-based diet with an occasional consumption of meat and meat products of ≤50 g/d) or (b) maintained a vegan (V) diet (i.e., avoidance of any food of animal origin) or (c) maintained an omnivore (OMN) diet (i.e., mixed diet with regular consumption of meat and meat products of ≥170 g/d).
The consumption limits for meat and meat products for FX were derived from the lower consumption recommendations of the DGE (300-600 g/week) [12]. The minimum intake of meat and meat products for OMN was based on the average consumption in Germany in 2019 (62 kg/year/person) [20,39,40]. Furthermore, all participants had to adhere to their respective diet for at least 1 year and had to be healthy, nonsmokers, aged between 25 and 45 years and with a body mass index (BMI) between 20 and 28 kg/m 2 .
Interested subjects with consumption rates of meat and meat products from ≥50 g/d ≤ 170 g/d were not included to achieve a clear differentiation between the three groups.
Other exclusion criteria were acute infections, several chronic diseases, regular use of laxatives, drugs, alcohol or medication abuse.
If eligible, the participants were invited for a study visit. On the latter, anthropometrics was measured, fasted blood samples were taken and several questionnaires were filled out.
This study was conducted at the Institute of Food Science and Human Nutrition at Leibniz University Hanover, Germany, according to the guidelines of the Declaration of Helsinki. Ethical approval was provided by the Ethics Committee of the Medical Association of Lower Saxony (Hanover, Germany). The study was registered in the German Clinical Trial Register (DRKS 00019887). All subjects provided written informed consent.
Nutrients 2022, 14,3038 3 of 23 Participants were matched by age and gender within their respective group and across all three groups.

Anthropometrics, Blood Samples and Measurement of Blood Pressure
The height and weight of the participants were measured on the examination day. The waist and hip circumference were determined using a tape measure. BMI was calculated according to the standard formula [41,42]. The body composition and the basal metabolic rate (kcal/d) were assessed using multifrequency bioelectrical impedance analysis, according to the manufacturer's recommendations, by trained nutritionists of Nutriguard M (Data Input Company, Darmstadt, Germany).
After overnight fasting (≥12 h), a licensed doctor took blood samples, obtained by puncture of an arm vein from each participant. On the same day, all samples were deep frozen transferred (−18 • C) and determined in the accredited and certified laboratory of Clinical Chemistry, Hannover Medical School, Germany. Blood pressure was measured, according to the manufacturer's recommendations, by trained nutritionists using Visomat (UEBE, Medical Wertheim, Külsheim, Germany).

Physical Activity
Physical activity was assessed using a validated German Freiburger Questionnaire [43]. This questionnaire considers all health-relevant activities, from which the daily energy expenditure in kcal/day was estimated. Energy expenditure and the basal metabolic rate were added to determine total energy expenditure per day and participant.

Dietary Habits
Food intake rates were assessed using the validated FFQ of the RKI, Germany [44]. However, the large variety of plant-based alternative products has not yet been ascertained. Consequently, 28 additional questions covering plant-based alternative products-low or high processed-were added (Appendix A).
Furthermore, prior to the study visit, participants were asked to fulfill a three-day diary record over three consecutive days, including two weekdays and one weekend day. These records were checked by nutritionists for completeness, readability and plausibility. Ambiguities were clarified with the participants if necessary. Data from these dietary logs were processed using PRODI ® 6.12, Nutri-Science GmbH, Freiburg, Germany (Organizational software for nutrition counselling based on the German Federal Food Code 3.02) to derive the daily intake of macro-and micronutrients of each participant.

Development of HEI-Flex
The HEI-flex is based on the validated HEI-2015 [45][46][47][48] and was adapted via the following major points.

Integration of Various Official Dietary Recommendations
Ratios between intake and intake recommendation had to be formed to calculate the HEI-flex score values for each component. Therefore, the German nutritional guidelines of DGE were used for FX and OMN, and, in addition, recommendations of the Giessen Vegan Pyramid for V [12,50]. If no German recommendations were available (components 10 and 13), guidelines from WHO as well as the dietary guidelines for Americans, respectively, were considered (Table 1) [26,51].
As Table 1 shows, the ratio of the mean daily intake and the official intake recommendation [12, 50,51] was calculated using the following formula for all components, except components (11) high-energy density foods (sweet) and (12) high-energy density foods (fat):

Using Individual Energy Expenditure Values
Regarding the calculations of foods listed in components 11 and 12 (foods with highenergy density), limits of total energy expenditure in % were used in accordance with comparable HEI-2015 components ("Added Sugars" and "Saturated Fats") [47,52]. Therefore, the energy content of the foods consumed within these components was determined and was put in relation to the individual total energy expenditure of each participant (Section 2.5.4 and Table 1).

Calculating Principles and Scoring
In order to be able to make an evaluation of the ratio of the mean daily intake and the official intake recommendation, four calculation principles were applied for the scoring of all components: the adequacy principles I and II and the moderation principles III and IV (Appendix B). Which principle was used depended on whether the food group components had a recommended or tolerated intake, according to the dietary recommendation.
The adequacy principle I was applied for foods listed in components 1-3; its recommendations focus on the minimal amount that should be consumed daily. Point losses occurred proportionally if the consumption rate was too low up to the minimum required intake amount.
The adequacy principle II was used to calculate the components where a consumption range was given (components 4-9 and 12). Point losses occurred proportionally if the supply quantity was too low, and if it was higher than the maximum quantity.
The moderation principle III was applied for components 13 and 14, as it is recommended to consume as little as possible. No consumption of foods from these components resulted in the maximum value (V max = 100 SP). The higher the consumption rates, the higher the point losses proportionally.
The moderation principle IV was used for components 10 and 11 as a tolerated consumption range is used. Point losses occurred proportionally when the tolerated maximum quantity was exceeded.
All components within the HEI-flex were weighted equally because proportional point losses were recorded by under-and overconsumption via the calculating principles mentioned above.
If a component contains different subgroups, the consumption rate of each subgroup was individually compared with the respective recommendations and SP were calculated. The average SP value of all subgroups forms the SP for the parent component (cf. Table 1).
Finally, a maximum of 100 points could be earned per component. Hence, a maximum of 1400 points were achievable for 14 components, which was divided by 14 to determine the total score points (SP). Thus, the HEI-flex scales from 0 to 100 SP, whereby a higher value reflects a higher adherence of official dietary recommendations.

Statistical Evaluation
Results are presented as median (25th, 75th percentile) or mean ± standard deviation (SD). The normal distribution of data was initially checked visually and subsequently with the Kolmogorov-Smirnov test. The one-way analysis of variance (ANOVA) was used with normal distribution to assess differences between the three diets. If the data set was not normally distributed, the Kruskal-Wallis test was used. The post hoc test with Bonferroni correction was performed if there were significant differences. The chi-square test was used to compare frequencies between the participants. The Mann-Whitney U test was used for calculations conducted between men and women.
Correlations between HEI-flex data and values from the food diaries were calculated using Spearman's Rho test. Moreover, p-values ≤ 0.05 were interpreted as statistically significant in this study.

Characterization of the Study Population
A total of 32 FX, 33 V and 29 OMN were included in the study ( Table 2). The participants did not differ significantly regarding age or sex distribution. The duration of adhering to the respective diet differed greatly, however, the minimum duration was at least 1 year as defined by the inclusion criteria. Significant differences also existed in the BMI among all study participants across the three diets (p = 0.005). As Table 2 shows, FX had a lower BMI in kg/m 2 when compared to OMN (FX 22 (21)(22)(23)(24)(25) vs. OMN 25 (23)(24)(25)(26)(27); p = 0.003), yet the BMI of V vs. OMN or V vs. FX was not significantly different. Comparing the energy expenditure of health-related activities (kcal/week), FX had neither higher values than OMN nor less than V, and it was not significant, respectively. However, the calorie requirement of V was more than twice as high as OMN (V 5793 (3111-8168) vs. OMN 2681 (1385-4286); p = 0.002).
In addition, no significant differences were found in "Beverages", "Carbohydrate sources", "Whole meal", "Nuts and seeds", "Processed meat and plant-based meat alternatives", "Drinks with high-energy density" and "Fats and oils and plant-based fat substitutes" between the three groups.

Indications for Relative and Construct Validity: Associations between Nutrient Intake of the 3-Day Food Record and HEI-Flex Calculations
In order to gain an initial insight into the relative and construct validity of the HEI-flex, the score components were correlated with the macronutrient intake based on the three-day food record (Table 4) and compared across the thirds of HEI-flex, moreover, cardiovascular risk parameters were also considered ( Table 5).  Data are shown as mean ± SD (normally distributed) or median with 25th, 75th percentile (not normally distributed). Differences between the thirds of the total study population regarding macronutrient intake and cardiovascular risk parameters were analyzed using ANOVA or the Kruskal-Wallis test with post hoc Bonferroni correction (α adj = 0.0167). p ≤ 0.05 = significant; n.s. = not significant. p-values in bold represent statistical significance. a = significant difference between the thirds "middle" and "upper". b = significant difference between the thirds "low" and "upper". HEI-flex = Healthy Eating Index-flexible.
Observing the first indications of relative validity, significant correlations were found between total score points of HEI-flex and the intake of "Total energy", "Total fat" and "Water", "Dietary Fiber", "Sodium", "Alcohol" with the Spearman's rank correlation coefficient (rs) (p rs < 0.05 respectively). The directions of these correlations implicated that higher HEI-flex SP are associated with a more beneficial macronutrient profile, i.e., a higher intake of dietary fiber and lower intake of total energy, fat, alcohol and sodium was correlated with higher diet quality according to the HEI-flex Furthermore, "Dietary Fiber" correlated positively with the component "Protein Sources", which is partly due to the fact that plant, as well as animal based protein sources, were captured in this component.
No significant correlations could be observed between the macronutrients "Protein", "Total fat" and "Carbohydrates" and SP of high-energy density foods and drinks because protein-, fat-and carbohydrate-rich foods went into different components of HEI-flex according to the processing or preparation method. For example, carbohydrate-rich foods from the food diary log went in different proportions into the components of "Whole meal", as well as "Carbohydrate sources" or "High-energy density foods (sweet). Similar, protein-rich foods were assigned to either the component "Foods with high-energy density -fat-" or "Protein sources".
To sum up, many rs-values ranged between 0.2 and 0.3. In addition, regarding first indications in construct validity, participants in the upper third of the HEI-flex had a higher intake of dietary fiber (p ≤ 0.001), but lower intake of energy, total fat, salt and alcohol. In the lower third, it was nearly inverse: there was a higher intake of energy, carbohydrates, total fat, salt and alcohol, but a lower intake of dietary fiber and water.
Regarding cardiovascular risk parameters, measured values showed the most positive tendencies within the upper third. For example, triglycerides, LDL-cholesterin and systolic blood pressure lowered with the higher third (n.s.), respectively. The decreasing insulin level with higher thirds were significant (p = 0.013).

Discussion
The aim of the present study was to develop an adapted HEI-2015, which enables the comparison of the diet quality of FX, V and OMN in a healthy German adult population. Therefore, a new HEI, termed HEI-flex, was developed integrating the wide range of plant-based alternatives, either low or high processed, and specific official V recommendations [12,50]. Thereby, the higher the compliance with the dietary guidelines, the higher the score points (SP) given (V max = 100 SP). In summary, FX scored lower than V but higher than OMN (FX 54 SP ± 8 vs. V 61 SP ± 10 vs. OMN 47 SP ± 9); p ≤ 0.001).
These HEI-flex results confirm previous studies observing the higher diet quality in plant-based diets compared to OMN, based on various diet quality indices. However, many of these diet quality indices (e.g., diet quality index, Mediterranean diet score, plant-based index) often differ in terms of their research aim (e.g., associations between dietary intake and health outcomes) or scoring methodology (e.g., allocation of foods only according to positive and negative) [53][54][55][56][57][58][59][60]. Therefore, only studies based on the original HEI versions are included in the discussion below.
Several studies have determined the diet quality for different target groups (e.g., national consumer surveys, children or seniors), based on an HEI [27,37,38,[61][62][63][64]. However, calculations of plant-based diets using an HEI are rare. Semi-vegetarians and V were considered in a Belgium study based on HEI-2010 only once [53]. Nutritional quality has already been determined in Germany several times based on an HEI [35][36][37][38], but so far, neither FX nor V have been explicitly studied regarding their diet quality. Furthermore, specific recommendations that apply to V have not yet been addressed in current HEI evaluations [35,37,38,50]. Therefore, the HEI-flex has integrated some important extensions and amendments, which are discussed below.

Development of HEI-Flex
Four key aspects are particularly focused on in the following concerning the HEI-flex development.
Firstly, the HEI-flex was based on FFQ data, including low or high processed plantbased alternatives. Thereby, the increasing consumption of these plant-based alternatives and their health quality were reflected. However, these foods have not been explicitly recorded yet [28,[36][37][38]52,53,64] or have only been added to protein sources without considering the degree of processing [53].
Secondly, the components of HEI-flex considered all food groups from the official German intake recommendations for OMN (including FX) and V [12,50]. Hence, countryspecific dietary habits and the particular nutritional recommendations of V have been taken into account. By contrast, other German HEI versions do not differentiate between various diets [36,38] or they integrated only a few specific German foods (e.g., whole meal bread) into the HEI-2015 component classification [37].
Thirdly, it is important to emphasize that all components in the present study were assigned the same weighting corresponding to the HEI-DEGS [38]. However, the HEI-flex differs from the original HEI-2015 and adapted German versions, which assign different maximum values to each component [35,52]. By comparison, every component in the HEI-flex may reach a maximum of 100 SP, whilst overconsumption of components with a consumption margin will be assessed with point losses. In this way, a much more differentiated assessment of the consumption rates becomes possible.
Lastly, while all other comparable HEI versions used only official standard values for total energy expenditure for calculating the scores of food components with high-energy density (sweet and fat) [37,38,52,53], the HEI-flex calculations in the present study consider the individual total energy expenditure of each participant assessed by physical activity questionnaires (cf. 2.3).
To sum up, the new developed HEI-FX integrates plant-based alternatives accounts for the processing of foods, considers specific recommendations for omnivores, as well as for vegans and uses the individual energy expenditure of each participant for calculating components with high energy density. This approach allows an evaluation of nutritional quality that can be applied to FX as well as to V and OMN (cf. Appendix C).

Components of HEI-Flex
The HEI-flex contains a special compilation of food components that focused especially on the wide range of plant-based alternatives and official V recommendations.
The components of the HEI-flex will be discussed in the following, pointing to conspicuous values: "Processed meat and plant-based meat alternatives" and "Foods with high density (sweet and fat)".
It is noteworthy that a high consumption of "Processed meat and plant-based meat alternatives" (e.g., cold cuts or sausage, processed plant-based burger patties or meat alternatives) were observed across all three diets in the present study. About a quarter of all participants achieved no SP in this component, which probably also lowered the total SP. The comparable consumption rates of processed meat, fish or sausage in SP results were also roughly found in the HEI-EPIC study, although the allocation of foods into components differed partly, and no distinction between diets was made [35]. Nevertheless, V has also only moderately low SP in this component compared to FX and OMN based on HEI-flex, emphasizing that highly processed animal-based products are a crucial component in the diet quality evaluation. In terms of health, these products are associated with negative effects on cardiovascular and carcinogenic risk [65,66]. Whether high processed plant-based alternatives have similar detrimental effects on cardiovascular health as high processed animal-based products still needs to be confirmed. Furthermore, very low SP were observed in the two components in the present study with "High-energy density foods". The overconsumption in these components has partly contributed to the lower total HEI-flex SP in FX and OMN, compared to V. Whereby, V also achieved a low SP in both of these components (≤35 SP). However, point losses in these components generally differ from previous studies [35,38,53,55]. Perhaps the associated food groups were not asked about in such detail or were assigned to other components in the other surveys mentioned above.
Finally, regarding the SP of all fourteen HEI-flex components, it is noticeable that FX did not score significantly higher than V or significantly lower than OMN in any component. These results are essentially in line with the Belgium study of Clarys et al. [53]. This is interesting when considering that neither the number of components matches the allocation of the plant-based alternatives nor the component-weighting and valuation principles. Hence, HEI-flex calculations and Clarys et al. (2014) show that an animal-reduced diet is associated with higher SP values.

Total Scoring with HEI-Flex
The total SP of FX, V and OMN in the present study were generally similar to the results observed by Clarys et al., who also reported higher SP for V when compared to semi-vegetarians and OMN based on HEI-2010 [53].
However, the total SP of this study were generally slightly higher, compared to the present study. On the one hand, these differences could be due to the subdivision of low or high processed foods into different components in HEI-flex, and, on the other hand, soft drinks with a high-energy density and alcohol were also included in the calculations in contrast to the Belgium study. Point losses in these components were probably one reason that the total SP of HEI-flex were significantly lower than other calculations based on HEI-2010, respectively, HEI-2015 [27,33,37,38,53,61]. Moreover, the majority of participants in Clarys et al. (2014) were female-the gender ratio in the present study was balanced. The higher SP of female compared to male study participants was also found in other comparable German studies based on an HEI [35,37].
Interestingly, the SP values of OMN Japanese women and men were quite similar to OMN American women and men based on HEI-2015 [61]. However, the individual SP of the components between the two populations were considerably different, presumably due to the country-specific foods, yet total scores were nearly the same. Equally, a similar interindividual variation was observed in the present study, since the standard derivations were quite wide in several components (e.g., Nuts and seeds, FX, 74 SP (14-100)).
In summary, national dietary habits, various food components, their individual weighting, different valuation principles and resulting maximum SP make a comparison of various HEI versions difficult. Moreover, due to the aspects already mentioned, a direct SP comparison does not appear to be purposeful and should be weighed up on a case-by-case basis.

Strengths and Limitations of the Study
The present study showed that the HEI-flex is a useful tool to evaluate the diet quality of diverse long-term dietary patterns by combining recommendations of mixed diets (FX and OMN) and V, considering the intake of plant-based alternatives and high processed foods.
Furthermore, the HEI-flex calculations ensure that the lack of consumption of certain food groups, for example, dairy products, and meat or plant-based alternatives, did not result in any disadvantages in scoring. However, under-or overconsumption in certain components also led to point losses. Another advance of the HEI-flex compared to the HEI-2015 is the inclusion of beverages and alcohol ("Beverages", "Drinks with high-energy density" and "Alcohol") in three independent components, which ensured a more differentiated assessment of diet quality. In addition, the HEI-flex can also calculate SP by gender (cf. Appendix C).
A limitation of the study is its cross-sectional design, which restricts conclusions regarding causality. Moreover, most of the participants were recruited via notices and online communities dealing with different diets. For this reason, a special health awareness of some participants cannot be excluded. Furthermore, the number of participants (n = 94) was relatively low as this study was considered to be an exploratory pilot study. The low number of participants could also be the reason for the predominantly weak correlations of relative validity in the present study. However, according to [67], low correlations (rs ≥ 0.2 ≤ 0.3) could already be an indication of significant values for sample sizes with n ≤ 250. Furthermore, HEI-flex calculations based on FFQ data, which did not distinguish between rice and whole meal rice or pasta and whole meal pasta, means that the results in this component may be partially inaccurate. Additionally, low processed meat, fish and vegetable protein were combined into one single component: "Protein sources", regardless of their nutritional-physiological value. The same applies to the foods that are listed in "Fats and oils and plant-based fat substitutes". Presumably, more differentiated queries, for example, according to brand names or manufacturers, would allow a more accurate assignment of the foods to the HEI-flex components and, thus, more precise results.

Future Research
The present results give first indications of the current dietary quality of FX compared to V and OMN in the German adult population-with FX performing better than OMN but less well than V. Further studies with a larger number of participants and detailed food consumption queries are needed to obtain valid data on dietary quality based on HEI-flex. In addition, the score needs to be validated to ensure resilient reliability in relative and construct validity.
Future prospective studies could also investigate whether associations exist between HEI-flex results and health benefits, for example, through extensive blood analyses. In addition, the inclusion of further sustainability criteria, for example, life cycle assessments, would be a possibility to evaluate diet quality more holistically in the future.

Conclusions
The HEI-flex enables the first comparison of the nutritional quality between the current common diets of FX, V and OMN in Germany. These results showed that a reduction in meat and meat products seems to be associated with a higher diet quality. However, the partly excessive consumption of processed products, animal-and plant-based, and the high consumption of foods with high-energy density was noticeable in large sections of the study participants. These aspects should be given more attention in the future, especially in the context of possible health consequences [66,[68][69][70][71][72][73].

Moderation Principles
Moderation Appendix C Table A2. Essential similarities and differences of HEI-flex and HEI-2015.
Assessment principles of the HEI-flex components; HEI-flex = Healthy Eating Index-flexible. ↑ points increase; ↓ points decrease. Own compilation of HEI-flex aspects and according to [52]. DGE= Deutsche Gesellschaft für Ernährung (German Nutrition Society). DGA = Dietary guidelines for Americans. WHO = World Health Organization. V min = maximum value. V max = minimum value. SP = Score Points.